The present invention relates in general to a solid-state image sensor, and more particularly, to a highly sensitive solid-state image sensor which is applied to a television camera system.
A solid-state image sensor such as a charge-coupled device (CCD) is small, light-weight and highly reliable as compared with a conventional image-sensing tube. Further, the solid-state image sensor also has an advantage in its extremely small production of image distortion or after-image. The solid-state image sensor can be applied widely in an industrial television camera, a home-use video camera or an electronic still camera due to such an advantage.
The desire for improving the sensitivity of this solid-state image sensor is endless. Further enhancement in the sensitivity of the solid-state image sensor will be increasingly desired in the future. The increasing sensitivity may be performed in general by enhancing the efficiency of the photoelectric conversion of the sensor itself and of the signal-to-noise ratio (SN ratio) of the sensing image by decreasing the noise components generated from the sensor itself. Particularly, the generation of noise components by the sensor itself is a main factor in determining the image sensitivity of the CCD. In other words, the suppression of the generation of noise in the sensor itself is the most effective means for improving the sensitivity of the sensor.
Various efforts have heretofore been directed to a technique for suppressing the generation of the noise in a solid-state image sensor to improve its sensitivity. However, it is still difficult to uniformly and preferably suppress or prevent noise from being generated over a wide frequency band.
For example, the conventional image-sensing system includes a signal output circuit for a charge-coupled device (CCD) well-known by those skilled in the art as a "floating diffusion type output circuit". According to the conventional image-sensing system, signal carriers photoelectrically converted by the CCD are sequentially transferred in the surface area of a substrate under the transfer electrodes of the CCD, and then flow to an electrically floating diffusion layer through an output gate electrode. The variation in the voltage of this layer is detected by a sense amplifier, and is output from the output terminal of the CCD. In the meantime, a pulse signal having a predetermined fixed period is applied to a reset gate electrode, and the voltage of the diffusion layer is reset to a reference voltage.
Noise components are generated during when the reset gate electrode is ON. Since the generation of this type of noise is caused by thermal noise in the MOSFETs, the noise is indispensable and cannot be prevented in advance. (The thermal noise is created by the fact that a current flows to the MOSFET formed of the reset gate electrode, reset drain electrode and n-diffusion layer.) Further, part of the reset pulse signal is mixed with a sense amplifier line through a capacitance between the reset gate electrode and the n-diffusion layer during the ON period of the reset gate electrode. If the amplitude of such a mixed pulse is varied, noise arises. Noise generated by such a mechanism in a signal output section is generally called "reset noise". When the output signal level from the solid-state image sensor is averaged by a known low-pass filter to increase the effective signal period, the above-described reset noise is mixed in the pure image signal (effective image signal) as low frequency components. As a result, the level of the output image signal is varied, and the SN ratio of the image signal thus obtained is decreased, thereby largely degrading the sensitivity of the CCD. Since the reset noise has low frequency components, this noise generates a very rough noise image on the reproduced image, thereby remarkably deteriorating the quality of the reproduced image.
The so-called "correlated double sampling processor" for improving the S/N ratio of the image signal by removing the reset noise of this type is well-known by those skilled in the art. According to this technique, a reference voltage in the stable period of the output signal of the CCD is forcibly clamped to a predetermined voltage, and the signal voltage during the effective signal period is then sampled. Further, an improved method for performing correlated double sampling after increasing the period by sampling the stable period once is proposed in Japanese Patent Disclosure (KOKAI) No. 55-163693.
However, according to the above-described prior technique, in a high speed image-sensing system in which the frequency of a clock pulse signal for driving a solid-state image sensor is set to several megahertz or higher, its stable period becomes short, and no margin exists in the clamping time, with the result that there arises no expectation in the improvements in the SN ratio. More concretely, a CCD which was designed and produced as a test model by the inventors and which has 400 horizontal picture elements and 500 vertical picture elements, had a horizontal clock frequency of 7.16 MHz. In this case, one period of the signal becomes 140 nsec. In order to operate reliably in this period, the pulse of the reset gate electrode RS is on approximately 35 nsec., the stable period becomes approximately 35 nsec., and the effective signal period becomes approximately 70 nsec. Though the stable period is 35 nsec., the period in which the reference voltage VA becomes sufficiently stable is actually approximately 15 nsec. Therefore, it is necessary to clamp within the 15 nsec., which is very difficult to do. In addition, since a narrow pulse having a pulse width of 15 nsec. or shorter is necessary, this pulse is mixed in the signal, and the SN ratio is deteriorated. Further, the correlated double sampling processor does not accurately operate if a high frequency noise is contained during the stable period of the output signal of the CCD, and the SN ratio is rather deteriorated from the previous state before clamping. If the clamped pulse contains pulse components in the signal band, the components are mixed as fixed noise with the output signal, and the reproduced image is remarkably deteriorated.
As described above, because the noise of the output circuit of the conventional CCD operates using at least several megahertz of the clock frequency, it has not heretofore been able to be removed; the SN ratio of the CCD is deteriorated; and a television of high sensitivity cannot be produced.